Information Technology Reference
In-Depth Information
5
Condition for Gradient Filter
The condition for determining the amplitude of the gradient filters is obtained next.
The following expression represents the gradient of function k
(
x
,
y
)
in direction
of
θ
.
cos
k
θ
x +
sin
θ
(
x
,
y
)
(38)
y
In the Fourier domain, this equation is transformed to
2
π
i
(
u cos
θ +
v sin
θ )
K
(
u
,
v
) .
(39)
are gradient filters in the 0 ,
60 and 120 direc-
As h a (
x
,
y
)
, h b (
x
,
y
)
and h c (
x
,
y
)
tions, the following conditions hold:
H a (
u
,
v
)=
2
π
iu
(40)
i 1
v
3
2
H b (
u
,
v
)=
2
π
2 u
+
(41)
i
v
3
2
1
2 u
H c (
u
,
v
)=
2
π
+
.
(42)
It thus follows that
m , n H mn
(
u
,
v
)=
2
π
iu
(43)
a
i 1
v
3
2
m , n H mn
(
u
,
v
)=
2
π
2 u
+
(44)
b
i
v
3
2
1
2 u
m , n H mn
(
u
,
v
)=
2
π
+
,
(45)
c
must hold for any u and v . By taking the limit as
0 and using first-order
approximations for the trigonometric functions, makes it possible to rewrite these
expressions as
|
u
|, |
v
|→
m , n a mn π mu = iu
4 i
(46)
3
2
3
2
1
2 u
1
2 u
m , n a mn π m (
4 i
+
v
)=
2
π
i
(
+
v
)
(47)
3
2
3
2
1
2 u
1
2 u
m , n a mn π m (
4 i
+
v
)=
2
π
i
(
+
v
) .
(48)
Search WWH ::




Custom Search